Long term roadmap

Test writing a huge amount of data, like 100 GB
Test creating a huge amount of open streams
Test horizontal scalability by measuring throughput while ramping up node count to ten-twenty?
Create power off simulation torture test suite
Special WAL optimized for our append-only use case
HA mode with Raft + BoltDB

Short-term TODO

Have subdir structure for storages as not to have too many files in one dir
Previous block sha256
Stats about stream (# of lines, # of bytes etc.)
"Training wheels"? i.e. separate append-only log for backup until we trust the mechanics of this as working?
Remove panic()s
Rename chunk -> block